Update: Google Secures Pentagon's Secret AI Contract, Security Terms Deemed Non-legally Binding

According to Dynamic Beating monitoring, The Information exclusively reported that Google has signed an agreement with the US Department of Defense, allowing the military to use Google's AI for classified work, with the term “any and all lawful government purposes.” The contract was amended from a nonclassified agreement in November last year, as confirmed by a Google spokesperson.

The contract includes a statement regarding a security baseline: both parties agree that the AI “is not intended for, and should not be used for,” large-scale domestic surveillance or autonomous weapons control. However, Charlie Bullock, Senior Researcher at the Law and AI Research Institute, bluntly stated that “is not intended for” has no legal force, merely indicating that both parties consider such use inappropriate but not constituting a breach. The contract also states that “this agreement does not give the right to control or veto legitimate government action decisions.”

Compared to the agreement signed with OpenAI in February, Google's terms are more lenient. OpenAI retained “full discretion over the security system,” while Google agreed to assist in adjusting AI security settings and filters at the government's request. A Google spokesperson stated that these filters are designed for consumers, and Google usually makes adjustments for enterprise customers. Google is the third company to sign a classified AI agreement with the Pentagon, following xAI and OpenAI. Previously, Anthropic was deemed a “supply chain risk” by the Pentagon and is in court battles for refusing to relax security restrictions.